Professional Documents
Culture Documents
M OBILE GIS is the extension of a geographic dissemination with large amount of geographic data [5].
Previously, data collection and editing process in the field
information system (GIS) which has been developed
take a lot of time and susceptible result some errors. GIS user
then can be run in the field. Previously GIS can only run in
must visit to the field for take the geographical data in the
the office with a desktop GIS. With mobile GIS, the user can
structure of paper maps. Then they carry out field edit using
retrieve, transfer, update, manipulate, analyze and display draft and notes on paper maps and structure. After they got the
geographic information everywhere and every time. The geographical data, the processes of field edits in the office
standard technologies that integrate in Mobile GIS application must be interpreted and manually entered into spatial
are wireless network for transfer and access data, mobile database. Consequently, the geographical data rarely accurate
devices for run GIS application in everywhere, and GPS and up-to-date as it could have been.
(Global Positioning System) to detect the location. With the rapid Mobile GIS developments now a day, GIS
can be taken into the field using some devices like compact
digital maps, laptop (mobile computer), PDA (Personal
Digital Assistant), etc. This new GIS technology enables
Authors are with Faculty of Computer Science and Information System,
company or GIS user to get real time geographic information,
University Technology of Malaysia, 81310 Skudai, Malaysia (e-mails:
muserie@gmail.com, shafry@utm.my, zaha@utm.my, daut@utm.my). update data much faster to their database and applications,
249
World Academy of Science, Engineering and Technology 22 2008
used together for index the data in one single spatial database R1 R4
3
250
World Academy of Science, Engineering and Technology 22 2008
multi-dimensional index tree, in a wireless broadcast clutter. The inability to easily locate the desired object and
environment in terms of latency and tuning time as well as apply the appropriate normalizations, are efficiently overcome
memory usage has also been proposed [15]. by the recursive quadtree procedure [18]. Fluid flow solver
based on unstructured quadtree / octree has been explored
III. QUADTREE SPATIAL DATA INDEXING OVERVIEW also. This indexing type Eulerian grids is coupled with a
A quadtree is a tree data structure in which each internal spectral Finite Element (p-FEM) structural mechanics solver
node has up to four children. Quadtrees are most often used to based on a Lagrangian description can predict bidirectional
partition a two dimensional space by recursively subdividing fluid-structure interaction (FSI) [19]. The new structure of
it into four quadrants or regions. The regions may be square or Multiversion Linear Quadtree (MVLQ) has been present
rectangular, or may have arbitrary shapes. A similar based on spatio-temporal access method. In this indexing
partitioning is also known as a Q-tree. Quadtrees has two structure can be used as an index mechanism for storing and
features which are: (1) they decompose space into adaptable accessing evolving raster images [20].
cells and (2) each cell (or bucket) has a maximum capacity. A novel fractal image coding based on Quadtree partition of
When maximum capacity is reached, the bucket splits, and the the adaptive threshold value has been proposed to show better
tree directory follows the spatial decomposition of the performance including the improved quality of the decoded
Quadtree [16]. image, shorter compression time and higher compression ratio
To cover the geometry, Quadtree spatial data indexing uses with another spatial indexing structure [21]. An alternative
fixed-size tiles controlled by tile resolution. Each tile has a non-pointer quadtree node codification to manage
fixed-size and shape, because the coordinate space has been geographical spatial data also has been presented. In this
decomposed a specific number of times. Then the tessellation structure, new codification is based on a variable sequence of
will be terminated. However all of this process could be z-ordered base four digits [22]. A new quadtree-based
happened if the resolution is the sole controlling factor [17]. decomposition of a polygon possibly with holes also has been
As we can see in Fig. 2, the tiles can be sorted linearly and presented. An approximate this indexing structure is good for
systematically at the exact level. Then it will visit the tiles ray shooting in the average case as defined by Aronov and
which determined by a space-filling curve. The unique Fortune [23].
numeric identifiers are also assigned to the tiles. MPI collective algorithm selection process and explore the
applicability of the quadtree encoding method has been
proposed. This method constructs quadtrees with different
properties from the measured algorithm performance data and
analyzes the quality and performance of decision functions
generated from these trees. Selecting the close-to-optimal
collective algorithm based on the parameters of the collective
call at run time is an important step in achieving good
performance of MPI applications [24]. The matrix quadtree
also has been presented based on the weighted in the form of
modification. The technique of distance transform is extended
to the weighted regions and applied to the robot path planning
[25]. A new data dissemination protocol that exploits
“Quadtree-based network space partitioning” to provide more
efficient routing among multiple mobile stimuli and sink
nodes also has been proposed. Because of a common
hierarchy of cluster-head nodes is constructed where the data
delivery to mobile sinks is independent of the current position
of mobile stimuli [26]. A 3D mesh is generated by using the
original quadtree triangulation also algorithm has been
proposed with a nontrivial algorithm for quadtree
triangulation. The aim of the study was to increase the
accuracy of a terrain triangulation while maintaining or
Fig. 2 Quadtree Algorithm Schema reducing the number of triangles [27].
A new data dissemination protocol that exploits “Quadtree-
Many researchers had explored about quadtree spatial based network space partitioning” to provide more efficient
indexing, even though quadtree is not recommended to use for routing among multiple mobile stimuli and sink nodes has
particular spatial indexing compare to R-tree spatial data been proposed. Simulation results show that in that work
indexing. However, there are many advantages using quadtree significantly reduces average energy consumption while
spatial indexing in the special situation. An algorithm based maintaining comparably higher data delivery ratio [28]. A
on applying eigenspace methods has been presented to a novel computational paradigm for detection of blood vessels
quadtree of related set images to solve the pose estimation in fundus images based on RGB components and quadtree
problem in the presence of occlusion and/or background decomposition also has been proposed. The proposed
algorithm employs median filtering, quadtree decomposition,
251
World Academy of Science, Engineering and Technology 22 2008
post filtration of detected edges, and morphological analysis of the data and operations indexing must be made in
reconstruction on retinal images [29]. A distributed quadtree advanced. The default of spatial index in oracle is an R-tree
index that adapts the MX-CIF quadtree was described that which the dimension of the data is two. If the spatial index has
enables more powerful accesses to data in P2P networks. This been built in two dimensions of data or more, only one built-
index has been implemented for various prototype P2P up operator (SDO_FILTER, the primary filter or index only in
applications [30]. the query) can be used in this situation. This operator
The coordinate space (for the layer where all geometric considers in all of dimension. Here we explore some of
objects are located) is issued to a procedure called tessellation comparisons between R-tree and Quadtree spatial data
in the linear quadtree indexing scheme of oracle spatial 10g, indexing.
which delineates elite and extensive cover tiles for all of pile The method to find the interiors of convex polygons is
up geometry. Tessellation is done by crumbling the coordinate different and can’t be applied to concave geometries [39].
space in a standard hierarchical mode. The range of Different technique should be used in the concave geometries
coordinates, the coordinate space, is viewed as a rectangle. At to get the interior estimations for concave query windows.
the first level of decomposition, the rectangle is divided into One of the methods is to break up recursively the MBR
halves along each coordinate dimension, producing four tiles. (Minimum Bounding Rectangle) of the concave geometry into
Each tile that cooperates with the geometry being tessellated is quadrants four times. This idea also can be applied to carry
further decomposed into four tiles. This procedure continues out a level four tilling via the MBR as the tilling domain. Then
until some termination criteria, such as size of the tiles or the the geometries are recognized along with the tiles that wrap
maximum number of tiles to cover the geometry is met. The the geometry. Aspirant MBRs are also tiled and the tiles for
results of the tessellation process on geometry are stored in a the aspirant MBR are seek out along with the core tiles for the
table, referred to as the SDOINDEX table [17]. concave query window. Current outcome [39] demonstrates
Spatial quadtree indexing utilizes fixed-size tiles to cover that this two-sided method for applying the intermediary filter
geometry. Fixed-size tiles are managed by tile resolution. If in R-trees attains extensive enhancements in query
the resolution is the single controlling reason, then tessellation performance. The core area detained using this method could
terminates when the coordinate space has been crumbled an be larger than that for some Quadtree indexes assembled using
exact number of times. Consequently, each tile is of a fixed a tiling level of 14 since the tiling domain in that case is the
amount and form. Fixed-size tile resolution is managed by a entire data space while the tiling domain for the R-tree is
user-selectable keyword named sdo_level. Smaller fixed-size simply the MBR of the concave geometry. Since this situation
tiles give better geometry approximations. happened, then the approximation of geometries cannot be
fine-tuned.
IV. CONSIDERATION OF CHOOSING R- TREE OR QUADTREE Quadtree indexing in Oracle Spatial database use two
SPATIAL DATA INDEXING IN ONE SINGLE SPATIAL strategies for index the geographical data. The first one is user
DATABASE SYSTEM choose and specified the tilling level of Quadtree indexing
R-tree and Quadtree indexes that use extensible framework method. The second one is approximate every geometry data
are the best spatial data indexing method from existing spatial thus the covering tiles can be minimize. At the appropriate
indexing method for indexing low-dimensional spatial data. stage the tile-code associating each tile using z-ordering.
Oracle spatial users can choose between one of two spatial Tessellation process has a responsibility to divide the tiles for
indexes for indexing low-dimensional spatial data: a linear geometry into internal and external part based on whether or
Quadtree [31] or an R-tree [32, 33, 34, 35, 36, 37]. For not they are fully interior to the geometry. The spatial_index
processing the queries, R-tree approach may be more efficient table receives all the wrapping tiles for geometry data beside
due to better maintenance of spatial immediacy but may be geometry rowid. The geometry data as a result of this method
slow in updates or index creation and implements its own may have numerous rows in the spatial_index table where
concurrency protocols on top of table-level concurrency every dissimilar tile-code, rowed, and internal or external
mechanisms. Since R-tree is built logically as a tree and status of the geometry stored in a row. Consequently geometry
physically using tables inside the database and the search approximation can be fine-tuned by setting the tilling level
involves recursive SQL for traversing the tree from root to and number of tiles. The last stage is speeding up queries for
relevant leaves. The linear Quadtree give results in simpler construct B-tree index on the tile_code, rowed, and status.
index creation, faster updates and inheriting of builtin B-tree R-tree has easier method in creating the index and tuning
concurrency control protocols. Since these indexes computes rather than Quadtree spatial index creation. Because in Oracle,
tile approximations for geometries and uses existing Btree R-tree will create the index by default without identify any
indexes for performing spatial search and other DML parameter. Moreover, setting the appropriate tuning parameter
operations. values of Quadtree can affect performance significantly. In
Oracle spatial provides spatial indexing to speed up access order to optimize performance, Quadtrees necessitate to be
and operation spatial data. That spatial index can be an R-tree fine-tuned by choosing a fitting tiling level. The storage and
or Quadtree index, or both. Every type of indexes can give the index creation costs rise as the tiling level increases.
good quality outcome in different situations. R-tree and Quadtree require tilling level more than R-tree, as a result the
Quadtree index also can be used together at the same time in storage that is needed to create the Quadtree index is bigger
one geometry column. To choose the appropriate indexing, than using R-tree index.
252
World Academy of Science, Engineering and Technology 22 2008
To recognize the nearest neighbors for a geometry, Linear referencing is an ordinary and suitable means to
SDO_NN (nearest neighbor) queries utilize the spatial index. correlate attributes or events to locations or portions of a
The parameter for tuning the parameter and may affect linear element. It has been extensively used in transportation
query’s performance utilize SDO_BATCH_SIZE. The applications (such as for highways, railroads, and transit
SDO_BATCH_SIZE only can be used if SDO_NN utilize an routes) and efficacies applications (such as for gas and oil
R-tree index to execute the procedure. Additionally this pipelines). The major benefit of linear referencing is its
keyword can’t be used when Quadtree index combined. The potentiality of locating attributes and events along a linear
value for this keyword will be affecting the performance of element with only one constraint (usually known as quantify)
nearest neighbor queries. instead of two (such as latitude/longitude or x/y in Cartesian
In Oracle Spatial, R-trees preserve their rational tree space). Sections of a linear feature can be referenced and
formation and are executed as a table where each node of the generated dynamically by signifying the start and end
R-tree communicates to a row in the table and a child pointer locations along the element without explicitly storing them. If
in the R-tree communicates to the row id of child row in the LRS (Linear Referencing System) data is indexed using a
same table. The root rowed (pointer) of the R-tree is stored in spatial quadtree index, only the first two dimensions are
the metadata for the index and allows routing from the root of indexed; the measure dimension and its values are not
the R-tree to the leaf nodes. Leaf nodes of the R-tree indexed.
accumulate one MBR for each geometry data together with SDO_WITHIN_DISTANCE utilizes the spatial index to
the geometry row id. Queries and updates attain the root of the recognize the set of spatial objects that are within some
R-tree and navigate down the tree to the leaves. particular distance of a given object (such as an area of
More details on this implementation can be obtained from interest or point of interest). Distance between two extensive
[38]. Note that each time an R-tree node is called; the objects (nonpoint objects such as lines and polygons) is
matching row from the spatial_index is chosen using a SQL described as the minimum distance between these two objects.
declaration internally. This means query and update If this operator is used with geodetic data, the data must be
processing in R-trees occupies processing of more recursive indexed with an R-tree spatial index. If this operator is used
SQL declarations than in the case of Quadtrees. As a result, with geodetic data and if the R-tree spatial index is created
some DML processes particularly update are likely to be more with 'geodetic=false' recognized, we cannot use the part
costly. And the approximation of geometries cannot be fine- parameter.
tuned. Geodetic data consists of geometries that have geodetic
In the case of Quadtrees, internal tiles for data geometries SDO_SRID values, reflecting the actuality that they are stand
are recognized at index creation time. Every geometry tiles are on a geodetic coordinate system (such as using longitude and
labeled with a "status" representing whether they are internal latitude) as contrast to a flat or estimated plane coordinate
tiles or external tiles. The status is set to boundary when any scheme. Thus indexing geodetic data will improve
element of the boundary of geometry touches a tile. significantly spatial features. However, quadtree indexes
Otherwise, if the tile is totally inside the geometry, the status cannot be geodetic indexes. If we generate a non-geodetic
is set to "interior". These status tags along with the tile-codes index on geodetic data, we cannot apply the part parameter
are evaluated with those of the query tiles at query time to with the SDO_WITHIN_DISTANCE operator.
execute the transitional filtering. Note that such labeling of the Distance, region, and angular units are completely
tiles that cover query and data geometries inflicts very little supported location data to be stored in the spatial database
transparency (one byte per row) on storage space, and does using the whole-earth geodetic form which guarantees
not change query or update performance. capacities across the earth’s exterior will be very precise.
A spatial R-tree index can index spatial data until four Location data can be stored in the database using the whole-
dimensions unlike the Quadtree that only can index only two earth geodetic model An R-tree index is required for a whole-
dimensions data. Each geometry data approximated with R- Earth index. In another hand a quadtree index cannot be used
tree index by a particular rectangle that minimally attaches the for a whole-Earth index.
geometry using MBR. R-tree index consists of a hierarchical To summarize all of those comparisons between R-tree and
index on the MBRs of the geometries for a sheet of geometries Quadtree spatial data indexing, below is the table for
in the layer. In the linear quadtree indexing design, the consideration to choose R-tree or Quadtree spatial data
coordinate space is focused to a procedure called tessellation, indexing in one spatial database system.
which describes restrictedly and extensively cover tiles for
every stored geometry data. The coordinate space in a
standard hierarchical approach is tessellated by
decomposition. At the first level of disintegration, the
rectangle is divided into halves along each coordinate
dimension creating four tiles. Each tile that interrelates with
the geometry being tessellated is further decomposed into four
tiles. This procedure keeps on until some termination
condition, such as size of the tiles or the highest quantity of
tiles to cover the geometry, is met.
253
World Academy of Science, Engineering and Technology 22 2008
254
World Academy of Science, Engineering and Technology 22 2008
255
World Academy of Science, Engineering and Technology 22 2008
the data that use an R-tree Spatial Indexing is faster than the TABLE III
R- TREE AND Q UADTREE E XPERIMENTS R ESULT
data that doesn’t use Spatial Indexing.
Number No Index Quadtree No Index R-tree
of (Seconds) Index (Seconds) Index
Records - (Seconds) – (Seconds)
Quadtree R-tree
1000 1.4 1.2 1.4 1.3
5000 1.6 1.6 1.6 1.5
10000 2.5 2.4 2.5 2.3
15000 3 2.8 3 2.7
20000 3.5 3.5 3.5 3
25000 5 3.9 5 3.6
30000 6 4.3 6 4
35000 8 4.6 8 4.4
VII. CONCLUSION
Fig. 4 R-tree Average Response Time Database tuning is an important step to solve system
bottleneck problem. The use of indexes is an efficient ways of
The result for Quadtree spatial index is depicted in Fig. 6, retrieving data from spatial database. R-tree and Quadtree
which is plotted the results of average response time in Mobile spatial index can be use as spatial database tuning methods to
GIS application. The line with blue color corresponds to the enhance Mobile GIS performance. MapViewer also one of
time measurement from the data that use a quadtree Spatial web base applications that suitable, fast, and has a good
Indexing, while the line with red color represents the time feature for mobile GIS applications to show geographical
measurement from the data that doesn’t use Spatial Indexing. map. Oracle Application Server and Oracle Spatial 10g have
We observe that the time saved increases as the database size function for web application server and recommended place to
expands. For the number of records, over 30,000, the time store and retrieve spatial data. The goals are to fine-tune the
required in the data that use a quadtree Spatial Indexing is original database for mobile users with a limited
almost two times more than the data that doesn’t use Spatial communication bandwidth, to improve the response time, and
Indexing. It is noted, however, that this is related to the actual show geographical map with a good feature. Further research
time of the query, since the database volume is varied as time will develop and test this combination of R-tree and Quadtree
changes. The response time is slightly different to the result spatial data indexing in one spatial database system from those
shown in figure (Fig. 5). Nevertheless, we can easily discover of the considerations.
that the time response on the data that use a quadtree Spatial
Indexing is faster than the data that doesn’t use Spatial ACKNOWLEDGMENT
Indexing. The authors would like to thank Ministry of Science,
Technology and Innovation Malaysia (MOSTI) and Universiti
Teknologi Malaysia (UTM) for their financial support under
e-science fund vot no. 79233.
REFERENCES
[1] Rajinder, S. N. (2004). Cartographic visualisation for mobile
application. Master’s thesis, ITC/IIRS.
[2] Peng, Z. R. & Tsou, M. H. (2003). Internet GIS: Distributed Geographic
Information Services for the Internet and Wireless Network. John Wiley
and Sons Inc.
[3] Mensah, E. (2007). Designing a Prototype Mobile GIS to Support
Cadastral Data Collection in Ghana, 44.
[4] Vckovski, A. (1999). Interoperability and spacial information theory.
Interoperating Geographic Information Systems.
[5] Kraak, M. J. (2002). Current trends in visualisation of geographic data
with special reference to cartography. Invited paper: In Proceedings of
Fig. 5 Quadtree Average Response Time the XXIIth INCA Congress 2002, Indian National Cartographic
Association: Convergence of Imagery Information and Maps, volume
22, 319–324.
Below is the table (Table III) to make it clear to see the [6] Chen, K., & Shi, W. (2002), A Study of Dynamic Database in Mobile
result. GIS.
[7] Guttman, A. (1984). R-Trees: A Dynamic Index Structure for Spatial
Searching. Proc. 1984 ACM SIGMOD International Conference on
Management of Data, pp. 47-57. ISBN 0-89791-128-8.
[8] Zhu, Q., Gong, J., Zhang, J. (2007). An efficient 3D R-tree spatial index
method for virtual geographic environments. ISPRS Journal of
256
World Academy of Science, Engineering and Technology 22 2008
Photogrammetry and Remote Sensing, Volume 62, Issue 3, August [32] A. Guttman (1984). R-trees: A dynamic index structure for spatial
2007, pp 217-224. searching. Proc. A CM SIGMOD Int. Conf. on Management of Data,
[9] Lee, T., Moon, B., Lee, S. (2006). Bulk insertion for R-trees by seeded pages 47-57.
clustering. Data & Knowledge Engineering, Volume 59, Issue 1, [33] T. Sellis, N. Roussopoulos, and C. Faloutsos (1988). The r+-tree: A
October 2006, pp 86-106. dynamic index for multi-dimensional objects. Procdf the Int. Conf. on
[10] Lee, M. L., Hsu, W., Jensen, C. S., Cui, B., Teo, K. L. (2003). Very Large Data Bases, 13:507-518.
Supporting Frequent Updates in R-Trees: A Bottom-Up Approach. [34] N. Beckmann, H. Kriegel, R. Schneider, B. Seeger (1990). The R* tree:
Proceedings 2003 VLDB Conference, 2003, pp 608-619. An efficient and robust access method for points and rectangles. In Proc.
[11] An, N., Kothuri, R. K. V., Ravada, S. (2003). Improving Performance ACM SIGMOD Int. Conf. on Management of Data, pages 322-331.
with Bulk-Inserts in Oracle R-Trees. Proceedings 2003 VLDB [35] S. Berchtold, D. A. Keim, and H. P. Kreigel (1996). The X-tree: An
Conference, 2003, pp 948-951. index structure for high dimensional data. Procff the Int. Conf. on Very
[12] Chan, E. P. F., & Chow, K. K. W. (2002). On multi-scale display of Large Data Bases.
geometric objects. Data & Knowledge Engineering, Volume 40, Issue 1, [36] S. T. Leutenegger, M. A. Lopez, J. M. Edgington (1997). STR: A simple
January 2002, pp 91-119. and efficient algorithm for R-tree packing. In Proe. Int. Conf. on Data
[13] Yun, J. K., Kim, D. O., Hong, D. S., Kim, M. H., Han, K.J. (2006). A Engineering.
real-time mobile GIS based on the HBR-tree next term for location [37] V. Gaede & O. Gunther (1998). Multidimensional access methods. ACM
based services. Computers & Industrial Engineering, Volume 51, Issue Computing Surveys, 30(2).
1, September 2006, pp 58-71. [38] K. V. Ravi Kanth, Siva Ravada, J. Sharma, J. Banerjee (1999). Indexing
[14] Francis, D. H., Madria, S., Sabharwal, C. (2008). A scalable constraint- medium-dimensionality data in oracle. In Proc. ACM SIGMOD Int.
based Q-hash indexing for moving objects. Information Sciences, Conf. on Management of Data.
Volume 178, Issue 6, 15 March 2008, pp 1442-1460. [39] K. V. Ravi Kanth & Siva Ravada (2001). Efficient processing of large
[15] Liu, C. M., & Fu, S. Y. (2008). Effective protocols for kNN search on spatial queries using interior approximations. In Symposium on Spatial
broadcast multi-dimensional index trees. Information Systems, Volume and Temporal Databases (SSTD).
33, Issue 1, March 2008, pp 18-35. [40] Oracle Application Server 10g White Paper (2007). Oracle Application
[16] Finkel, R., & Bentley, J. L. (1974). Quad Trees: A Data Structure for Server 10g.
Retrieval on Composite Keys. Acta Informatica 4 (1): 1-9.
[17] Oracle Spatial 10g White Paper (2006). Oracle Spatial Quadtree
Indexing, 10g Release 1 (10.1). Maruto Masserie Sardadi obtained his B.Sc. degree
[18] Chang, C. Y., Maciejewski, A. A., Balakrishnan, V., Roberts, R. G., in Computer Science from the University of
Saitwal, K. (2006). Quadtree-based eigen decomposition for pose Indonesia (UI), Indonesia in 2004. All of his working
estimation in the presence of occlusion and background clutter. experienced is in IT area, some in oil industry. He is
[19] Geller, S., Talke, J., Krafczyk, M. (2007). Lattice-Boltzmann Method on currently a Computer Science M.Sc. research student
Quadtree-Type Grids for Fluid Structure Interaction. Fluid-Structure in University Technology of Malaysia (UTM). His
Interaction, 270-293. research interests include database, GIS, and
[20] Tzouramanis, T., Vassilakopoulos, M., Manolopoulos, Y. (2000). petroleum.
Multiversion Linear Quadtree for Spatio-Temporal Data. Current Issues
in Databases and Information Systems, 279-292.
[21] Zhang, L. & Xi, L. F. (2007). A Novel Fractal Image Coding Based on
Mohd Shafry Mohd Rahim obtained his B.Sc.
Quadtree Partition of the Adaptive Threshold Value. Theoretical
degree in Computer Education, 1999 and M.Sc.
Advances and Applications of Fuzzy Logic and Soft Computing, 504-
degree in Computer Graphics, 2002 from University
512.
Technology of Malaysia (UTM). He is currently a
[22] Varas, J. (2007). Mobile Robot Path Planning Among Weighted Regions
Computer Science PhD. research student in
Using Quadtree Representations. Computer Aided Systems Theory -
University Putra of Malaysia (UPM) and UTM
EUROCAST’99, 239-249.
Computer Science lecturer. His research interests
[23] Cheng, S. W. & Lee, K. H. (2008). Quadtree Decomposition, Steiner
include GIS and spatiotemporal database.
Triangulation, and Ray Shooting. Algorithms and Computation, 368-
377.
[24] Grbovic, J. P., Fagg, G. E., Angskun, T., Bosilca, G., Dongarra, J. J.
(2006). MPI Collective Algorithm Selection and Quadtree Encoding. Zahabidin Jupri obtained his B.Sc. degree in
Recent Advances in Parallel Virtual Machine and Message Passing Mathematics Education from University of Malaya
Interface, 40-48. (UM) and M.Sc. degree in Info. Mgmt. from
[25] Varas, J. (2007). Mobile Robot Path Planning Among Weighted Regions Sheffield, UK. He is currently UTM Computer
Using Quadtree Representations. Computer Aided Systems Theory - Science lecturer.
EUROCAST’99, 239-249.
[26] Mir, Z. H. & Ko, Y. B. (2006). A Quadtree-Based Data Dissemination
Protocol for Wireless Sensor Networks with Mobile Sinks. Personal
Wireless Communications, 447-458.
[27] Samet, R. & Ozsavas, E. (2007). Optimization of Quadtree
Triangulation for Terrain Models. Advanced Concepts for Intelligent
Vision Systems, 48-59. Daut Daman, obtained his B.Sc. degree in Math.
[28] Mir, Z. H. & Ko, Y. B. (2007). A quadtree-based hierarchical data Computer Science, 1979 from University Sains of
dissemination for mobile sensor networks. Telecommunication Malaysia (USM) and M.Sc. degree in Applied
Systems, 117-128. Computing, 1984 from Cranfield Institute of
[29] Reza, A. W., Eswaran, C., Hati, S. (2007). Diabetic Retinopathy: A Technology, UK. He is currently a Head Department
Quadtree Based Blood Vessel Detection Algorithm Using RGB of Computer Graphics and Multimedia Faculty of
Components in Fundus Images. Journal of Medical Systems, 147-155. Computer Science UTM.
[30] Tanin, E., Harwood, A., Samet, H. (2006). Using a distributed quadtree
index in peer-to-peer networks. The VLDB Journal, The International
Journal on Very Large Data Bases, 165-178.
[31] H. Samet (1989). The design and analysis of spatial data structures.
Addison-Wesley Publishing Co..
257